Health burden attributable to ambient PM2.5 in China

Environmental Pollution 223 (2017) 575e586

Contents lists available at ScienceDirect

Environmental Pollution journal homepage: www.elsevier.com/locate/envpol

Health burden attributable to ambient PM2.5 in China* Congbo Song a, b, Jianjun He c, **, Lin Wu a, b, Taosheng Jin a, e, Xi Chen d, Ruipeng Li a, b, Peipei Ren a, b, Li Zhang a, b, Hongjun Mao a, b, * a

College of Environmental Science & Engineering, Nankai University, Tianjin, 300071, China Center for Urban Transport Emission Research, Nankai University, Tianjin, 300071, China c State Key Laboratory of Severe Weather & Key Laboratory of Atmospheric Chemistry of CMA, Chinese Academy of Meteorological Sciences, Beijing, 100081, China d Department of Occupational and Environmental Health, School of Public Health, Tianjin Medical University, Tianjin, 300070, China e State Environmental Protection Key Laboratory of Urban Particulate Air Pollution Prevention, Tianjin, 300071, China b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 17 October 2016 Received in revised form 11 January 2017 Accepted 20 January 2017 Available online 3 February 2017

In China, over 1.3 billion people have high health risks associated with exposure to ambient fine particulate matter (PM2.5) that exceeds the World Health Organization (WHO) Air Quality Guidelines (AQG). The PM2.5 mass concentrations from 1382 national air quality monitoring stations in 367 cities, between January 2014 and December 2016, were analyzed to estimate the health burden attributable to ambient PM2.5 across China. The integrated exposure-response model was applied to estimate the relative risks of disease-specific mortality. Disease-specific mortality baselines in province-level administrative units were adjusted by the national mortality baseline to better reveal the spatial inequality of the health burden associated with PM2.5. Our study suggested that PM2.5 in 2015 contributed as much as 40.3% to total stroke deaths, 33.1% to acute lower respiratory infection (ALRI, 2), were then removed from the hourly raw data. The missing data were filled with the time-interpolated method. The daily city-level pollution was represented by the daily average pollutant concentrations of all the monitoring stations within this city (Li et al., 2016b).

2.1.2. Population data The population data in this study was downloaded from the website of the National Bureau of Statistics of the People's Republic of China (http://www.stats.gov.cn/ztjc/zdtjgz/zgrkpc/dlcrkpc/). The latest census data were from the Sixth National Population Census carried out in 2010. Although some cities might have released their new version of population data updated after 2010, we still used the population data from the Sixth National Population Census to ensure data integrity and consistency, considering a few small cities did not have relatively correct population data since 2010. The resolution of the census data was on a county-level, which could be summed up to city-level for our study purpose.

2.1.3. Mortality baseline Cause-specific mortality data of China in 2015 were obtained from GBD 2015 database (Wang et al., 2016b). Province-level mortalities in 2013 from the GBD 2013 study (Zhou et al., 2016a) (Fig. S1) were adjusted by GBD 2015 database to derive provincelevel mortality data in 2015. The causes of mortality were defined by the International Classification of Diseases 10 (ICD-10) code: IHD (I20-I25); stroke (I60-I67, I69.0, I69.1, I69.2, I69.3); lung cancer (C33, C34); COPD (J40-J44); and ALRI (J09-J15.8, J16-J16.9, J20-J21.9, P23-P23.9).

C. Song et al. / Environmental Pollution 223 (2017) 575e586

2.2. Methods 2.2.1. Health impact estimation Health end points associated with PM2.5 exposure could be categorized into morbidity and chronic mortality. For chronic mortality relative risk, we adopted the global IER model (Table S1 and Fig. S2) (Burnett et al., 2014) with the counterfactual concentration of PM2.5 (where RR ¼ 1) in the range of 5.8e8.8 mg m3. For morbidity excess risk, we adopted the linear exposure-response functions (Table S2) (Xie et al., 2016b). The attributable fraction (AF) measures the contribution of a risk factor to disease or mortality (Ezzati et al., 2003). The following model was used to estimate the disease-specific AF associated with exposure to ambient PM2.5 (Anenberg et al., 2010; Lelieveld et al., 2013):

AFi;j ¼

RRCi ;j 1 RRCi ;j

(1)

where Ci is the annual mass concentration of PM2.5 at site i, and RRCi ;j is the relative risk for disease j at exposure level Ci calculated from IER functions. The attributable mortality rates (AMR) of total deaths combining IHD, stroke, COPD and LC were estimated as follows:

AMRi ¼

n X

AFi;j yi;j

(2)

j¼1

where yi;j is the mortality rate for disease j at site i. The AF of total deaths caused by IHD, stroke, COPD and LC were estimated as follows:

AMR AFi ¼ Pn i j¼1 yi;j

(3)

To evaluate the health benefits (HB) of population-weighted average (PWA) PM2.5 mass concentrations in China meeting the WHO interim targets (IT1, IT2 and IT3) and AQG, we used:

For mortality;

For morbidity;

n X RRCi ;j RRRef ;j HBi ¼ Popi yi;j RRCi ;j j¼1

HBi ¼ Popi

n X

!

ERFj Ci CRef

(4)

(5)

j¼1

where Pop, ERF are the exposed population number and exposureresponse functions. The suffixes i, j, Ref represent different cities i, different health endpoints j (for chronic mortality, they are IHD, stroke, COPD and LC, for morbidity, they are work loss days and morbidity cases of respiratory, cerebrovascular and cardiovascular hospital admission, chronic bronchitis, asthma attacks, respiratory symptom days), and reference concentrations (IT1: 35 mg m3, IT2: 25 mg m3, IT3: 15 mg m3 and AQG: 10 mg m3). The health benefits (HB) are equal to attributable mortalities (AM) when the reference concentrations are the counterfactual concentration of PM2.5 (where RR ¼ 1). 2.2.2. Reference scenario To estimate the PM2.5 mass concentrations at each NAQMS in different scenarios (year 2005 which is the 10-y time lag for this study, WHO IT1, IT2, IT3 and AQG), the annual PM2.5 standard scores (z-scores) at each NAQMS in 2015 were utilized to adjust the national annual PM2.5 concentrations (2005: 68.6 mg m3 (Kan

577

et al., 2012; Liu et al., 2017), IT1: 35 mg m3, IT2: 25 mg m3, IT3: 15 mg m3, AQG: 10 mg m3) to derive site-specific PM2.5 concentrations. 2.2.3. Statistical analysis We used a logistic regression to fit the city-level PM2.5 mass concentration (x), and AF (x) with an associated cumulative exposed population percentage (F(x)). The regression is defined as follows:

1 p

FðxÞ ¼ 1þ

(6)

x x0

where: p ¼ Hill's slope. The Hill's slope refers to the steepness of the curve and its dispersion or spread. x0 ¼ Inflection point, and the inflection point is defined as the point on the curve where the curvature changes direction. x0 is the exposed data that is covering 50% of the total population. 2.2.4. Uncertainty analysis 95% confidential intervals (CI) of attributable mortalities (AM) were given in this study. The average, 2.5%, and 97.5% cause-specific RR values under certain PM2.5 exposure concentration were given by Burnett et al. (2014). The fit parameters for the IER model were presented in Table S1. Considering the lag health effects of longterm PM2.5 exposure, we also conducted a 10-year time lag analysis using the PM2.5 concentration in 2005 and the mortality data in 2015. 3. Results 3.1. PM2.5 exposure assessment Fig. 1 (a) shows the spatial distribution of annual PM2.5 mass concentrations (mg m3) at NAQMS in 2014, 2015 and 2016. The annual mass concentrations of ground PM2.5 showed a significant geographic variation across China (from 17 mg m3 to 143 mg m3 in 2014, from 10 mg m3 to 131 mg m3 in 2015, from 8 mg m3 to 146 mg m3 in 2016). The spatial patterns of annual PM2.5 mass concentrations in China remained steady during this study periods. As illustrated in Fig. 1 (a), the regions with highest PM2.5 concentration were located in the Northern China Plain, the Middle-Lower Yangtze Plains, the Sichuan Basin and Tarim Basin. The Z-scores (Fig. S3) in 2015 showed a good agreement with those in 2014 (n ¼ 881, R2 ¼ 0.83) and 2016 (n ¼ 1494, R2 ¼ 0.82) (as shown in Fig. 1 (b)), suggesting that the spatial inequalities caused by the PM2.5 pollution remain steady recent years. Assuming the generally spatial-differentiation of PM2.5 pollution were not significantly altered, the z-scores of NAQMS in 2015 could be utilized to adjust the national annual PM2.5 reference concentrations of different scenarios (year 2005 which is the 10-y time lag for this study: 68.6 mg m3 (Kan et al., 2012; Liu et al., 2017), WHO IT1: 35 mg m3, IT2: 25 mg m3, IT3: 15 mg m3, AQG: 10 mg m3) to estimate corresponding site-specific annual PM2.5 mass concentrations. The density distributions of annual PM2.5 concentrations at each NAQMS of different reference scenarios are displayed in Fig. 2. PM2.5 pollution in China showed significant declines especially from 2014 (62.8 mg m3) to 2016 (48.1 mg m3). The spreads of the PM2.5 concentrations suggested that China still has a long way to go especially for those cities with annual PM2.5 concentration higher than 50 mg m3 if China aims to achieve WHO IT1 (or CAAQS II standard), IT2, IT3 and AQG standards. From the logistic population regression (Fig. 3), 50% population in China were exposed to annual PM2.5 mass concentration with

578


Fig. 1. Spatial distribution of annual PM2.5 mass concentration (a) and correlations of PM2.5 z-scores (b) from NAQMS in 2014, 2015 and 2016 (the longitude of NAQMS in 2014 and 2016 were shifted to left side (0.7 ) and right side (þ0.7 ) of those in 2015, respectively).

63.7 mg m3 in 2014, 53.0 mg m3 in 2015 and 48.2 mg m3 in 2016. In China, 95%, 87% and 81% population were exposed to PM2.5 concentrations higher than WHO IT1 (or CAAQS) standard in 2014, 2015 and 2016, respectively. None of the population of China lived in areas meeting the WHO guideline of 10 mg m3 during this study period. PM2.5 concentration in Beijing-Tianjin-Hebei (BTH), Yangtze River Delta (YRD), Pearl River Delta (PRD) and Sichuan-Chongqing (CY) city clusters all showed significant declines from 2014 to 2016. However, the entire population living in the four city clusters was still exposed to PM2.5 concentrations above WHO IT2, IT3 and AQG. The citizens living in the BTH city cluster were exposed to the highest PM2.5 concentration, with 91% (2014), 86% (2015) and 73% (2016) of the city-cluster's population exceeding 70 mg m3 (twice as much as the CAAQS of China). The PRD had the lowest annual

PM2.5 concentrations with nearly half population in PRD were exposed to PM2.5 concentrations meeting WHO IT1 (or CAAQS) standard. 3.2. Mortality attributable to PM2.5 City-level AFs of disease-specific mortality was estimated by applying IER functions to city-level annual PM2.5 mass concentrations. Fig. 4 illustrates the disease-specific AFs in China in 2005, 2014e2016, and scenarios of meeting WHO interim targets (IT1, IT2, and IT3). In 2015, the city-level AF (%) varied from 10.4 (2014: 17.1, 2016: 9.5) to 34.0 (34.7, 36.0) for IHD, 4.1 (17.9, 4.9) to 49.5 (49.7, 50.0) for stroke, 3.9 (8.4, 9.5) to 29.7 (31.0, 33.8) for COPD, 4.5 (10.4, 4.2) to 37.8 (39.4, 42.7) for LC, and 2.3 (9.7, 2.6) to 53.7 (55.4, 58.3) for ALRI.


579

Fig. 2. Density plots of site-specific annual PM2.5 in different scenarios. Lines represent smooth fit of density function.

Fig. 3. Population exposure to PM2.5 in four megacity clusters and China (the vertical ordinate is cumulative distribution of population percentage over the range of PM2.5 concentrations in the areas where the residents lived).

Fig. 4. Population exposure of disease-specific AFs in 2005, 2014e2016, and WHO interim targets.

580


Fig. 5. Attributable fractions (%) (a) and attributable mortality rates (105 y1) (b) combining IHD, stroke, LC, and COPD in 2015. (Circle for the main analysis, square for the 10-year time lag analysis, the offset of longitude on the map is 0.7 ).

In a previous global study (Burnett et al., 2014), the percent AF exposure varied among countries from 2 to 41 for IHD, 1 to 43 for stroke,